TPC Benchmark-H (TPC-H) is a database benchmark published by the Transaction Processing Performance Council (TPC). TPC-H is widely used in academia and various industries to evaluate the performance of decision support applications. TPC-H is an important guideline that is used to select databases.

AnalyticDB for PostgreSQL ranks first in the TPC-H 30 TB performance list after rigorous testing of large-scale analytical performance and atomicity, consistency, isolation, durability (ACID) properties of database transactions. AnalyticDB for PostgreSQL is a world-leading database service that is the first Chinese product to be listed among TPC-H 30 TB rankings for performance and cost efficiency.

2021052401

TPC-H consists of a suite of business-oriented ad-hoc queries and concurrent data modifications. TPC-H tests the transactional and analytical performance of databases.

In TPC-H, AnalyticDB for PostgreSQL faces the following challenges:

  • Large data sets of 30 TB: AnalyticDB for PostgreSQL uses data sets of up to 30 TB in size. The largest table can contain 180 billion rows. Large data volumes pose a major challenge to data import, storage, and computing performance.
  • Complex association analysis: Association analysis covers operations such as multi-table join, union query, multi-dimensional data filtering, and high-precision numerical calculation. Complex association analysis poses challenges to optimizers on how to optimize union queries, select the best join order, and select the distribution mode for tables. Complex association analysis also poses challenges to the compute engine on memory management, code generation, and execution scheduling.
  • Batch data writes in real time: Write and delete operations on hundreds of millions of data records pose challenges to the transaction processing performance and the column storage engine required by an availability and partition tolerance (AP) system.
  • High-throughput concurrent reads and writes: AnalyticDB for PostgreSQL uses data sets of up to 30 TB in size, which require at least 10 concurrent query streams and multiple refresh streams to perform multiple rounds of add and remove operations on a large amount of data. This poses challenges to the concurrency and ACID properties of the system.
  • Distributed transaction verification: TPC-H tests the transaction processing capabilities including ACID properties of the database system. In the tests, coordinator nodes and compute nodes are shut down. TPC-H tests the availability of the system in extreme cases such as power outage. TPC-H requires complex procedures, high system performance, and strict regulation. In some cases, devices must be restarted multiple times.